Global and local path planning in natural environment by physical modeling

Abstract

In the field of robotics, some autonomous vehicles work in natural sites. One problem for the autonomous system is to plan a path to the goal. A usual method uses potential fields to find a global path. With this method, obstacles must be clearly defined. But this is not the case in such natural environments. A new physically based potential method is introduced to find paths such as, for instance, a winding path to a pass. Moreover, locally, vehicles have to cope with a geometrically unstable environment and thus have a complex dynamic behavior. Thus it is necessary to elaborate a physical model of the robot-environment system. Generally, the two plannings, global and local, are independent. The introduced potential method resorts to the simulations of the vehicles to find a feasible path.